1 /* Operations with long integers.
2    Copyright (C) 2006-2016 Free Software Foundation, Inc.
3 
4 This file is part of GCC.
5 
6 GCC is free software; you can redistribute it and/or modify it
7 under the terms of the GNU General Public License as published by the
8 Free Software Foundation; either version 3, or (at your option) any
9 later version.
10 
11 GCC is distributed in the hope that it will be useful, but WITHOUT
12 ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
13 FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
14 for more details.
15 
16 You should have received a copy of the GNU General Public License
17 along with GCC; see the file COPYING3.  If not see
18 <http://www.gnu.org/licenses/>.  */
19 
20 #ifndef DOUBLE_INT_H
21 #define DOUBLE_INT_H
22 
23 /* A large integer is currently represented as a pair of HOST_WIDE_INTs.
24    It therefore represents a number with precision of
25    2 * HOST_BITS_PER_WIDE_INT bits (it is however possible that the
26    internal representation will change, if numbers with greater precision
27    are needed, so the users should not rely on it).  The representation does
28    not contain any information about signedness of the represented value, so
29    it can be used to represent both signed and unsigned numbers.  For
30    operations where the results depend on signedness (division, comparisons),
31    it must be specified separately.  For each such operation, there are three
32    versions of the function -- double_int_op, that takes an extra UNS argument
33    giving the signedness of the values, and double_int_sop and double_int_uop
34    that stand for its specializations for signed and unsigned values.
35 
36    You may also represent with numbers in smaller precision using double_int.
37    You however need to use double_int_ext (that fills in the bits of the
38    number over the prescribed precision with zeros or with the sign bit) before
39    operations that do not perform arithmetics modulo 2^precision (comparisons,
40    division), and possibly before storing the results, if you want to keep
41    them in some canonical form).  In general, the signedness of double_int_ext
42    should match the signedness of the operation.
43 
44    ??? The components of double_int differ in signedness mostly for
45    historical reasons (they replace an older structure used to represent
46    numbers with precision higher than HOST_WIDE_INT).  It might be less
47    confusing to have them both signed or both unsigned.  */
48 
49 struct double_int
50 {
51   /* Normally, we would define constructors to create instances.
52      Two things prevent us from doing so.
53      First, defining a constructor makes the class non-POD in C++03,
54      and we certainly want double_int to be a POD.
55      Second, the GCC conding conventions prefer explicit conversion,
56      and explicit conversion operators are not available until C++11.  */
57 
58   static double_int from_uhwi (unsigned HOST_WIDE_INT cst);
59   static double_int from_shwi (HOST_WIDE_INT cst);
60   static double_int from_pair (HOST_WIDE_INT high, unsigned HOST_WIDE_INT low);
61 
62   /* Construct from a fuffer of length LEN.  BUFFER will be read according
63      to byte endianess and word endianess.  */
64   static double_int from_buffer (const unsigned char *buffer, int len);
65 
66   /* No copy assignment operator or destructor to keep the type a POD.  */
67 
68   /* There are some special value-creation static member functions.  */
69 
70   static double_int mask (unsigned prec);
71   static double_int max_value (unsigned int prec, bool uns);
72   static double_int min_value (unsigned int prec, bool uns);
73 
74   /* The following functions are mutating operations.  */
75 
76   double_int &operator ++ (); // prefix
77   double_int &operator -- (); // prefix
78   double_int &operator *= (double_int);
79   double_int &operator += (double_int);
80   double_int &operator -= (double_int);
81   double_int &operator &= (double_int);
82   double_int &operator ^= (double_int);
83   double_int &operator |= (double_int);
84 
85   /* The following functions are non-mutating operations.  */
86 
87   /* Conversion functions.  */
88 
89   HOST_WIDE_INT to_shwi () const;
90   unsigned HOST_WIDE_INT to_uhwi () const;
91 
92   /* Conversion query functions.  */
93 
94   bool fits_uhwi () const;
95   bool fits_shwi () const;
96   bool fits_hwi (bool uns) const;
97 
98   /* Attribute query functions.  */
99 
100   int trailing_zeros () const;
101   int popcount () const;
102 
103   /* Arithmetic query operations.  */
104 
105   bool multiple_of (double_int, bool, double_int *) const;
106 
107   /* Arithmetic operation functions.  */
108 
109   /* The following operations perform arithmetics modulo 2^precision, so you
110      do not need to call .ext between them, even if you are representing
111      numbers with precision less than HOST_BITS_PER_DOUBLE_INT bits.  */
112 
113   double_int set_bit (unsigned) const;
114   double_int mul_with_sign (double_int, bool unsigned_p, bool *overflow) const;
115   double_int wide_mul_with_sign (double_int, bool unsigned_p,
116 				 double_int *higher, bool *overflow) const;
117   double_int add_with_sign (double_int, bool unsigned_p, bool *overflow) const;
118   double_int sub_with_overflow (double_int, bool *overflow) const;
119   double_int neg_with_overflow (bool *overflow) const;
120 
121   double_int operator * (double_int) const;
122   double_int operator + (double_int) const;
123   double_int operator - (double_int) const;
124   double_int operator - () const;
125   double_int operator ~ () const;
126   double_int operator & (double_int) const;
127   double_int operator | (double_int) const;
128   double_int operator ^ (double_int) const;
129   double_int and_not (double_int) const;
130 
131   double_int lshift (HOST_WIDE_INT count) const;
132   double_int lshift (HOST_WIDE_INT count, unsigned int prec, bool arith) const;
133   double_int rshift (HOST_WIDE_INT count) const;
134   double_int rshift (HOST_WIDE_INT count, unsigned int prec, bool arith) const;
135   double_int alshift (HOST_WIDE_INT count, unsigned int prec) const;
136   double_int arshift (HOST_WIDE_INT count, unsigned int prec) const;
137   double_int llshift (HOST_WIDE_INT count, unsigned int prec) const;
138   double_int lrshift (HOST_WIDE_INT count, unsigned int prec) const;
139   double_int lrotate (HOST_WIDE_INT count, unsigned int prec) const;
140   double_int rrotate (HOST_WIDE_INT count, unsigned int prec) const;
141 
142   /* You must ensure that double_int::ext is called on the operands
143      of the following operations, if the precision of the numbers
144      is less than HOST_BITS_PER_DOUBLE_INT bits.  */
145 
146   double_int div (double_int, bool, unsigned) const;
147   double_int sdiv (double_int, unsigned) const;
148   double_int udiv (double_int, unsigned) const;
149   double_int mod (double_int, bool, unsigned) const;
150   double_int smod (double_int, unsigned) const;
151   double_int umod (double_int, unsigned) const;
152   double_int divmod_with_overflow (double_int, bool, unsigned,
153 				   double_int *, bool *) const;
154   double_int divmod (double_int, bool, unsigned, double_int *) const;
155   double_int sdivmod (double_int, unsigned, double_int *) const;
156   double_int udivmod (double_int, unsigned, double_int *) const;
157 
158   /* Precision control functions.  */
159 
160   double_int ext (unsigned prec, bool uns) const;
161   double_int zext (unsigned prec) const;
162   double_int sext (unsigned prec) const;
163 
164   /* Comparative functions.  */
165 
166   bool is_zero () const;
167   bool is_one () const;
168   bool is_minus_one () const;
169   bool is_negative () const;
170 
171   int cmp (double_int b, bool uns) const;
172   int ucmp (double_int b) const;
173   int scmp (double_int b) const;
174 
175   bool ult (double_int b) const;
176   bool ule (double_int b) const;
177   bool ugt (double_int b) const;
178   bool slt (double_int b) const;
179   bool sle (double_int b) const;
180   bool sgt (double_int b) const;
181 
182   double_int max (double_int b, bool uns);
183   double_int smax (double_int b);
184   double_int umax (double_int b);
185 
186   double_int min (double_int b, bool uns);
187   double_int smin (double_int b);
188   double_int umin (double_int b);
189 
190   bool operator == (double_int cst2) const;
191   bool operator != (double_int cst2) const;
192 
193   /* Please migrate away from using these member variables publicly.  */
194 
195   unsigned HOST_WIDE_INT low;
196   HOST_WIDE_INT high;
197 
198 };
199 
200 #define HOST_BITS_PER_DOUBLE_INT (2 * HOST_BITS_PER_WIDE_INT)
201 
202 /* Constructors and conversions.  */
203 
204 /* Constructs double_int from integer CST.  The bits over the precision of
205    HOST_WIDE_INT are filled with the sign bit.  */
206 
207 inline double_int
from_shwi(HOST_WIDE_INT cst)208 double_int::from_shwi (HOST_WIDE_INT cst)
209 {
210   double_int r;
211   r.low = (unsigned HOST_WIDE_INT) cst;
212   r.high = cst < 0 ? -1 : 0;
213   return r;
214 }
215 
216 /* Some useful constants.  */
217 /* FIXME(crowl): Maybe remove after converting callers?
218    The problem is that a named constant would not be as optimizable,
219    while the functional syntax is more verbose.  */
220 
221 #define double_int_minus_one (double_int::from_shwi (-1))
222 #define double_int_zero (double_int::from_shwi (0))
223 #define double_int_one (double_int::from_shwi (1))
224 #define double_int_two (double_int::from_shwi (2))
225 #define double_int_ten (double_int::from_shwi (10))
226 
227 /* Constructs double_int from unsigned integer CST.  The bits over the
228    precision of HOST_WIDE_INT are filled with zeros.  */
229 
230 inline double_int
from_uhwi(unsigned HOST_WIDE_INT cst)231 double_int::from_uhwi (unsigned HOST_WIDE_INT cst)
232 {
233   double_int r;
234   r.low = cst;
235   r.high = 0;
236   return r;
237 }
238 
239 inline double_int
from_pair(HOST_WIDE_INT high,unsigned HOST_WIDE_INT low)240 double_int::from_pair (HOST_WIDE_INT high, unsigned HOST_WIDE_INT low)
241 {
242   double_int r;
243   r.low = low;
244   r.high = high;
245   return r;
246 }
247 
248 inline double_int &
249 double_int::operator ++ ()
250 {
251   *this += double_int_one;
252   return *this;
253 }
254 
255 inline double_int &
256 double_int::operator -- ()
257 {
258   *this -= double_int_one;
259   return *this;
260 }
261 
262 inline double_int &
263 double_int::operator &= (double_int b)
264 {
265   *this = *this & b;
266   return *this;
267 }
268 
269 inline double_int &
270 double_int::operator ^= (double_int b)
271 {
272   *this = *this ^ b;
273   return *this;
274 }
275 
276 inline double_int &
277 double_int::operator |= (double_int b)
278 {
279   *this = *this | b;
280   return *this;
281 }
282 
283 /* Returns value of CST as a signed number.  CST must satisfy
284    double_int::fits_signed.  */
285 
286 inline HOST_WIDE_INT
to_shwi()287 double_int::to_shwi () const
288 {
289   return (HOST_WIDE_INT) low;
290 }
291 
292 /* Returns value of CST as an unsigned number.  CST must satisfy
293    double_int::fits_unsigned.  */
294 
295 inline unsigned HOST_WIDE_INT
to_uhwi()296 double_int::to_uhwi () const
297 {
298   return low;
299 }
300 
301 /* Returns true if CST fits in unsigned HOST_WIDE_INT.  */
302 
303 inline bool
fits_uhwi()304 double_int::fits_uhwi () const
305 {
306   return high == 0;
307 }
308 
309 /* Logical operations.  */
310 
311 /* Returns ~A.  */
312 
313 inline double_int
314 double_int::operator ~ () const
315 {
316   double_int result;
317   result.low = ~low;
318   result.high = ~high;
319   return result;
320 }
321 
322 /* Returns A | B.  */
323 
324 inline double_int
325 double_int::operator | (double_int b) const
326 {
327   double_int result;
328   result.low = low | b.low;
329   result.high = high | b.high;
330   return result;
331 }
332 
333 /* Returns A & B.  */
334 
335 inline double_int
336 double_int::operator & (double_int b) const
337 {
338   double_int result;
339   result.low = low & b.low;
340   result.high = high & b.high;
341   return result;
342 }
343 
344 /* Returns A & ~B.  */
345 
346 inline double_int
and_not(double_int b)347 double_int::and_not (double_int b) const
348 {
349   double_int result;
350   result.low = low & ~b.low;
351   result.high = high & ~b.high;
352   return result;
353 }
354 
355 /* Returns A ^ B.  */
356 
357 inline double_int
358 double_int::operator ^ (double_int b) const
359 {
360   double_int result;
361   result.low = low ^ b.low;
362   result.high = high ^ b.high;
363   return result;
364 }
365 
366 void dump_double_int (FILE *, double_int, bool);
367 
368 #define ALL_ONES (~((unsigned HOST_WIDE_INT) 0))
369 
370 /* The operands of the following comparison functions must be processed
371    with double_int_ext, if their precision is less than
372    HOST_BITS_PER_DOUBLE_INT bits.  */
373 
374 /* Returns true if CST is zero.  */
375 
376 inline bool
is_zero()377 double_int::is_zero () const
378 {
379   return low == 0 && high == 0;
380 }
381 
382 /* Returns true if CST is one.  */
383 
384 inline bool
is_one()385 double_int::is_one () const
386 {
387   return low == 1 && high == 0;
388 }
389 
390 /* Returns true if CST is minus one.  */
391 
392 inline bool
is_minus_one()393 double_int::is_minus_one () const
394 {
395   return low == ALL_ONES && high == -1;
396 }
397 
398 /* Returns true if CST is negative.  */
399 
400 inline bool
is_negative()401 double_int::is_negative () const
402 {
403   return high < 0;
404 }
405 
406 /* Returns true if CST1 == CST2.  */
407 
408 inline bool
409 double_int::operator == (double_int cst2) const
410 {
411   return low == cst2.low && high == cst2.high;
412 }
413 
414 /* Returns true if CST1 != CST2.  */
415 
416 inline bool
417 double_int::operator != (double_int cst2) const
418 {
419   return low != cst2.low || high != cst2.high;
420 }
421 
422 /* Return number of set bits of CST.  */
423 
424 inline int
popcount()425 double_int::popcount () const
426 {
427   return popcount_hwi (high) + popcount_hwi (low);
428 }
429 
430 
431 #ifndef GENERATOR_FILE
432 /* Conversion to and from GMP integer representations.  */
433 
434 void mpz_set_double_int (mpz_t, double_int, bool);
435 double_int mpz_get_double_int (const_tree, mpz_t, bool);
436 #endif
437 
438 namespace wi
439 {
440   template <>
441   struct int_traits <double_int>
442   {
443     static const enum precision_type precision_type = CONST_PRECISION;
444     static const bool host_dependent_precision = true;
445     static const unsigned int precision = HOST_BITS_PER_DOUBLE_INT;
446     static unsigned int get_precision (const double_int &);
447     static wi::storage_ref decompose (HOST_WIDE_INT *, unsigned int,
448 				      const double_int &);
449   };
450 }
451 
452 inline unsigned int
453 wi::int_traits <double_int>::get_precision (const double_int &)
454 {
455   return precision;
456 }
457 
458 inline wi::storage_ref
459 wi::int_traits <double_int>::decompose (HOST_WIDE_INT *scratch, unsigned int p,
460 					const double_int &x)
461 {
462   gcc_checking_assert (precision == p);
463   scratch[0] = x.low;
464   if ((x.high == 0 && scratch[0] >= 0) || (x.high == -1 && scratch[0] < 0))
465     return wi::storage_ref (scratch, 1, precision);
466   scratch[1] = x.high;
467   return wi::storage_ref (scratch, 2, precision);
468 }
469 
470 #endif /* DOUBLE_INT_H */
471